Building presentation decks from your own numbers with AI

How the bot works

Nine steps. Pulling the reports, placing the numbers, laying out the slides and checking every figure against its source are ordinary code and cost almost nothing to run; the model is only asked to write a sentence about a movement it can point at.

1. It starts from the reports, not from a topic. The export from your accounting system, the KPI table, the ticket report, the production sheet, the occupancy list. A prompt describing what the deck should say is not a source.

2. It fills the deck you already agreed. The slide list, in your order, with your titles. The pack is a fixed shape that people have learned to read, and a bot that reorders it or adds a slide of its own has made the pack harder to use rather than easier.

3. Every number names where it came from. This figure is that report, that tab, that line, as of that cutoff. The citation lives with the deck, so a question in the meeting has an answer that does not require rebuilding the number.

4. It uses your definitions. What counts as revenue, which tickets are open, how utilization is measured, which period is the comparison. The definitions come from your own documentation, and a metric with two definitions in two systems is held rather than averaged.

5. The commentary comes from the movement. What changed against the prior period and against the target, written in your words, with the figure beside it. A sentence that cannot be tied to a number on the slide does not get written.

6. Anything it cannot source is a gap on the slide. A missing export, a figure nobody published this month, a metric that broke when a system changed. The slide says so, in place, rather than carrying a number that looks finished.

7. It asks why, instead of guessing why. A movement past the threshold you set comes back as a question for the person who owns that line. An explanation nobody recorded is not something a model should supply on their behalf in front of a board.

8. A person approves, gap list first. The draft deck waits with the gaps, the held metrics and the open questions on top. Somebody approves the version they read, and that version is the file that goes out. The bot never presents it and never sends it.

9. It tells you which numbers nobody can explain. The movements with no recorded reason, the metrics that disagree between systems, the slides that have been carrying a stale figure. That list is the useful output in the first month, because none of it is visible while the pack still looks finished.

What you configure

It can be configured to run in any business. This is what that actually means — the list of things that are yours, not ours.

Your deck

The slide list, the order, the titles and which slides are required every time. Changing the pack is a decision you make, not something a run does.

Your sources

Which report feeds which slide, where it comes from, how it is exported, and which source wins when two of them hold the same metric.

Your definitions

What each measure means in your business, written down once — the period, the inclusion rules, the comparison, the rounding.

Your periods and cutoffs

When the month closes for this pack, which figures are allowed to be provisional, and what happens when a source has not been updated yet.

Your commentary thresholds

How big a movement has to be before it earns a sentence, which lines always get one, and which slides are numbers only.

Your template

The layout, the fonts and colors, the chart styles, the footer, and the version of the template that is current.

Your audiences

The same numbers for a board, a client or an internal review — which slides each audience sees, and what never leaves the building.

Your approvers

Who approves the pack, who owns each line when a question goes back, and which audiences always need a named person.

Your release switches

Which packs may be assembled automatically on your calendar, which wait for somebody to ask, and whether the finished file is delivered anywhere at all. Delivery starts off.

Transparent, and configurable by you

The dark band below is what the system will not do. This is what you can see and change — and it is how we build everything, not something invented for this page.

You see what it did, and why

Every number on every slide names the report, tab and line behind it, every gap says what was missing, and every pack records the sources it read, the cutoff, the questions it raised and who approved it. A deck can be checked against your own reports rather than taken on trust.

Versioned prompts you can tune yourself

The slide list, the source map, the definitions, the commentary thresholds and the prompts behind writing a sentence about a movement are yours to tune. A change is a reviewable diff, replayed against packs you have already presented so you can see what it would have said differently, and rolled back the same way.

A human in the loop wherever judgment actually matters

Pulling the reports, placing the numbers, writing the sourced commentary and assembling the file are automatic. Deciding what the numbers mean, what to say about them, what the board is told and whether a figure is good enough to present are a person's — that boundary is a property of the system, not a promise about how carefully we'll use it.

Every run logged, along with what it cost

One row per pack and one per gap: the sources it read with their timestamps, every figure with its citation, every held metric, the questions it raised, who approved the file and where it went — plus token usage and cost for that run.

It can be configured to run in any business. The pack and the definitions are yours: your slides, your sources, your measures, your thresholds, your approvers. The decks on this page are one configuration of it, not the limit of it.

Works with your tools

This builds the pack in the format your audience already reads, from the systems you already run. We're not selling you a dashboard, and your reports stay the record.

• Microsoft 365 and PowerPoint
• Google Workspace and Slides
• Excel workbooks
• QuickBooks Online
• NetSuite
• A BI or dashboard tool
• A PSA or ticketing system
• A spreadsheet of KPIs

Works well for

• Nonprofits — The board packet is rebuilt by hand every month, out of the accounting system and a spreadsheet.
• Managed service providers — A quarterly review per client, the same slides every time, different numbers in each one.
• Property management companies — Owner reports go out per property, from a system that produces an export rather than a pack.
• Manufacturers — A monthly operations review with the same charts, pulled from line data and the ledger.
• School districts — Board reporting runs on a published schedule, with the same measures expected every time.

How a rollout works

Nothing is presented on day one and nothing is delivered anywhere. The first milestone is a pack you have already presented, rebuilt from your own reports, so you can compare it slide by slide with the one that went to the room.

1. 01Discovery

   The quote call, then a working session. We go through the pack as it exists, which report feeds which slide, how each measure is defined, your cutoffs, your template, who owns each line, who approves, and which audiences see what. You get the slide-to-source map, the definitions and a written scope.

2. 02Dry run

   A period you have already reported, rebuilt. Nothing is delivered. You compare every number with the pack that went out and read the gap list — and the gap list is where the work is, because a slide that has been carrying a stale figure looks exactly like one that has not.

3. 03Assisted mode

   Live on one pack. The bot assembles into your template, your owner reviews and approves, and the file goes out the way it always has. Who presents and who answers for the numbers does not change.

4. 04Optional limited automation

   Optional, and narrower than it sounds. A short list of packs you name, assembled on your calendar from sources that arrived on time, still unapproved and undelivered, with every hard stop in force and the kill switch. Plenty of teams stop at step 03, and that's a good place to stop.

What we don't claim

• No numbers we haven't earned. No hours-saved claim, no decks-per-month figure, no accuracy percentage. We have not built your pack. What a run takes and what it costs come out of the dry run, and they will be your numbers.
• No customer claims on this page. No names, no case studies, no logos.
• It is not a dashboard and it is not your data warehouse. If the numbers for a slide do not exist anywhere, assembling a deck will not create them. Making a measure exist and be trustworthy is data work, and it is a different engagement with a different shape.
• A deck that assembles itself does not make the meeting better. The value shows up when the pack is the same every month, the figures are traceable and the questions arrive before the meeting rather than during it. What anyone decides in the room is still a human matter.
• A gap on a slide is doing its job. It would be easy to make every slide look complete, and a slide that looks complete is how somebody ends up defending a number nobody can source. The design would rather show the hole.

Questions people ask

Can AI build our monthly deck?

Yes, from reports you already produce. An agent can pull the sources behind each slide, place every figure with its report and line named, write commentary from the movements it can trace, mark anything it cannot source as a gap, assemble the pack in your template and hold it for approval. What it will not do is invent a figure or add a slide of its own.

Where do the numbers come from?

From the reports you point it at - the accounting export, the KPI sheet, the ticket report, the production or property system - and each one is cited on the slide. A number with no source does not get placed, which is the whole difference between this and a deck generator that starts from a prompt.

Does it write the commentary?

It writes sentences about movements it can tie to a figure on the slide, in your words. A movement past your threshold with no recorded reason comes back as a question for the person who owns that line rather than as an explanation the bot composed.

Will it change our template or our slide order?

No. The pack is the slide list you agreed, in your order, in your template. Nothing is added, nothing is reordered and nothing is dropped because a source was missing - a missing source shows up as a gap on the slide that needed it.

What happens if a report is late?

The run comes back marked partial, with the slides that could not be filled named. It does not carry last month's figure forward, annualize around the hole or quietly leave the slide out.

Can it make different versions for different audiences?

Yes, and that is a configuration rather than a judgment: which slides a board sees, which a client sees, what never leaves the building. A variant nobody approved is a hard stop.

Does it send the deck to the board or the client?

Only if you turn that on, and only after a named person approves the exact file - delivery starts off. It never presents and never joins a meeting.

Is this the same as the notes and follow-ups from a meeting?

No, and they are deliberately separate. This builds the pack that goes into the meeting from reports. Turning what was said in a meeting into commitments and tasks is a different bot with its own page.

Is there a record of how a slide was built?

Yes, per pack: every source it read with its timestamp, every figure with its citation, the metrics it held, the questions it raised, who approved the file and where it went - plus the token usage and cost for that run.